簡易檢索 / 詳目顯示
| 研究生： |
宋承翰 Cheng-Han Song |
|---|---|
| 論文名稱： |
快速估算NPC行為—以撞球遊戲為例 Rapid NPC Behavior Estimation—A Case Study of Billiard Game |
| 指導教授： |
戴文凱
Wen-Kai Tai |
| 口試委員: |
鮑興國
Hsing-Kuo Pao 謝東儒 Tung-Ju Hsieh 鄭文皇 Wen-Huang Cheng |
| 學位類別： |
碩士 Master |
| 系所名稱： |
電資學院 - 資訊工程系 Department of Computer Science and Information Engineering |
| 論文出版年： | 2017 |
| 畢業學年度： | 105 |
| 語文別： | 中文 |
| 論文頁數： | 67 |
| 中文關鍵詞： | 撞球 、機器學習 、神經網路 |
| 外文關鍵詞： | NPC Behavior Estimation, Billiard, Neural Network |
| 相關次數： | 點閱：288 下載：6 |
| 分享至: |
| 查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
人工智慧應用於遊戲中的行為計算總需要經過繁複的設計與計算，若想要精準地得到其行為結果往往需要花費大量的時間；其時間無法使用多層級之計算方法，用以預測未來。因此本論文希望能提出一種快速估算NPC行為之方法，使多層級之計算時間不再是開發瓶頸；並以此研究結果進行進一步的AI架構開發。本論文將以撞球遊戲為例子，用以描述我們實際應用此方法之情形。
本論文透過撞球遊戲為例子，經由研究撞球中的力學意義，分解其目標之行為成多項可估算之小行為；接著依照其分解之多個小行為，設計不同的實驗以發掘藏在資料集之中的特徵；接著利用機器學習之方法將可估算之小行為訓練成多個估算模型；至此，所有估算之小行為模型皆已具備，再參照上述遊戲中的力學意義、不同特徵之相互關係，設計最終快速估算之架構，並將可估算之行為模型、力學公式等，代入快速估算之架構；最後在Unity軟體中將此架構移植回撞球遊戲中，驗證我們的方法之正確性。
由實驗結果得知，我們所提出的方法可以在快速的時間之內（與原本遊戲中的計算相比，快90~100倍），預估出撞球遊戲中母球與子球之行為，且母球估算的準確度落在十分精準的區間（撞擊一次子球平均誤差於0.9個球直徑）。
Artificial intelligence applied to the game in the calculation of the total need to go through the complicated design and calculation, if you want to accurately get their behavior results often need to spend a lot of time; its time cannot use multi-level calculation method to predict the future. Therefore, this paper hopes to propose a method to estimate the NPC behavior quickly, so that the multi-level computing time is no longer a development bottleneck; and further research on the results of AI architecture development. This paper will use the pool game as an example to describe the actual application of this method.
In this paper, the billiard game as an example, by studying the mechanics of the pool, the decomposition of the behavior of its target into a number of small behavior can be estimated; then in accordance with its decomposition of a number of small behavior, design different experiments to explore hidden features in the data set. And then use the method of machine learning to train the small behavior can be estimated into a number of estimation model; At this point, all the estimated small behavior models are available, and then refer to the above-mentioned game in the mechanical significance of the different characteristics of the relationship, Design the final rapid estimation of the structure, and the estimated behavior model, mechanical formula, into the rapid estimation of the structure. Finally in the Unity software will be transplanted back to the ball game, verify the correctness of our approach.
From the experimental results we know that the proposed method can be in the fast time (compared with the original calculation of 90 to 100 times faster), to predict the behavior for cue ball and object ball, and the accuracy of the estimation of the cue ball falls within a very precise interval (average error of 0.9 ball diameter).
BERNSTEIN, P. A., HADZILACOS, V. & GOODMAN, N. 1987. Rrency Control and Recovery in Database Systems, Addison− Wesley.
BLUETEAK. 2017. Unity-Neural-Network [Online]. Available: https://github.com/Blueteak/Unity-Neural-Network [Accessed].
CHURCHILL, D. & BURO, M. Hierarchical Portfolio Search: Prismata’s robust ai architecture for games with large search spaces. Proceedings of the Artificial Intelligence in Interactive Digital Entertainment Conference, 2015.
GAO, J., HE, Q., ZHAN, Z. & GAO, H. Dynamic modeling based on fuzzy Neural Network for a billiard robot. Networking, Sensing, and Control (ICNSC), 2016 IEEE 13th International Conference on, 2016. IEEE, 1-4.
GAO, J., ZHU, M., LIANG, H., GUO, X. & HE, Q. Design of the Multiple Neural Network Compensator for a Billiard Robot. Networking, Sensing and Control (ICNSC), 2015 IEEE 12th International Conference on, 2015. IEEE, 17-22.
HAGAN, M. T., DEMUTH, H. B. & BEALE, M. H. 1996. Neural network design, PWS Pub. Co., Boston, 3632.
HEATON, J. 2008. Introduction to neural networks with Java, Heaton Research, Inc.
HSU, Y.-W. 2010. The Simulation of Basic Billiard Techniques and The Analysis of Frictional Force.
INTERNATIONAL GAMES SYSTEM CO., L. 2017. Pool Ace - King of 8 Ball [Online]. Available: https://play.google.com/store/apps/details?id=com.igs.poolace&hl=en [Accessed].
LANDAU, L. & LIFSHITZ, E. 1976. Mechanics, vol. 1. Course of theoretical physics, 3.
MATHAVAN, S., JACKSON, M. R. & PARKIN, R. M. 2010. A theoretical analysis of billiard ball dynamics under cushion impacts. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 224, 1863-1873.
UNITY. 2017. Physics.Sphere.Cast [Online]. Available: https://docs.unity3d.com/ScriptReference/Physics.SphereCast.html [Accessed].
高家颖, 何秋阳 & 詹志新 2017. 基于神经网络的撞球机器人控制器设计. 北京航空航天大学学报, 533-543.
董增華 2012. 學撞球標準指南. 中, 職業選手密技大公開, 董增華.
戴文凱. 2017. 遊戲人工智慧與關卡難易度 [Online]. Taipei Game Developers Forum. Available: http://2017.tgdf.tw/speakers/wen-kai_tai/, http://gamelab.csie.ntust.edu.tw/public/news_files/TGDF-2017.pdf [Accessed 2017].
